Arao AI
Back to search results

DeepSeek-R1-Distill-Qwen-14B · IQ2_XS

Start a discussion
NVIDIA H2097GBllama.cpp v0.45L2 Cross-framework verified

This page aggregates 1 real-world runs of DeepSeek-R1-Distill-Qwen-14B (IQ2_XS) on NVIDIA H20 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

67.98 tok/s

Decode

Decode speed

225.25 tok/s

Prefill

Prefill speed

1.77 s

TTFT

Time to first token

16.21 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

17 days ago

Last verified

Performance

  1. llama.cpp · IQ2_XS (current)Decode 67.98 · Prefill 225.25 ·
  2. llama.cpp · Q4_K_M Decode 42.42 · Prefill 221.34 ·
  3. vLLM · Q2_K Decode 62.74 · Prefill 220.68 ·
  4. vLLM · Q6_K Decode 33.53 · Prefill 137.11 ·

Core figures

Decode (avg)
67.98 tok/s
Prefill (avg)
225.25 tok/s
TTFT (avg)
1.77 s
VRAM (avg)
16.21 GB
MTP acceptance rate
—
TTFB
0.54 GB
Power draw
339.3 W

Configuration

Member-level fields are taken from the most recent run

Model
DeepSeek-R1-Distill-Qwen-14B
Quantization
IQ2_XS
Framework
llama.cpp
Version
v0.45
Context length
8192 tokens
Batch size
256
GPU layers
33
Flash Attention
On

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
NVIDIA H20
Nominal VRAM
97 GB
Measured VRAM (avg)
16.21 GB
OS
Windows 11 24H2
Driver
565.57
CUDA
12.6
Power draw
339.3 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.45 · Windows 11 24H2 · CUDA 12.6 · 8192 ctx

    67.98 tok/s

    Decode

    225.25 tok/s

    Prefill

    1.77 s

    TTFT

    16.21 GB

    VRAM

    —

    MTP

    339.3 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=56 running model_id=71 at IQ2_XS